Ai2’s Molmo 2 open video models challenge Gemini and Qwen

The Allen Institute for AI (Ai2) has announced Molmo 2, an open multimodal model suite for video, image, and multi-image understanding with strong pointing, tracking, and reasoning capabilities. The release includes 4B, 8B, and fully open 7B variants plus nine new video-centric datasets, all available with weights and evaluation tools under open terms.